Bile Duct Involvement Portends Poor Prognosis in Resected Gallbladder Carcinoma

ABSTRACT

BACKGROUND:

Gallbladder cancer (GBC) carries an unfavorable prognosis with high mortality. This retrospective study was conducted to identify prognostic factors after resection of GBC, to assist in selecting appropriate surgical and adjuvant therapy.

METHODS:

Sixty-two patients from two institutions were identified with GBC by pathology. In 25, the cancer was unresectable at presentation. The remaining 37 patients comprised the study population. Log-rank analysis was used to assess univariate association with disease-free survival (DFS) and disease-specific survival (DSS). Cox regression was used for multivariate analysis.

RESULTS:

Median DFS and DSS were 22.6 and 28.5 months respectively, with a median follow-up of 44.2 months. On univariate analysis, bile duct (BD) involvement was significantly associated with decreased DFS (P ≤ .001) and DSS (P = .004). BD involvement was uniformly fatal. LN involvement was not significantly associated with DFS or DSS (P = .85, P = .54).

CONCLUSIONS:

All patients with BD involvement in our population died of the disease. The subset of patients with resectable GBC and BD involvement is a group that is at high risk for recurrence and should be treated as such. In our small population, preoperative and intraoperative methods evaluating BD involvement were unreliable.

Although gallbladder cancer (GBC) remains a relatively rare disease in North America, it continues to be the most common biliary tract malignancy and the eighth most common gastrointestinal (GI) malignancy in the United States. In addition, it has a high incidence in other areas of the world, including Chile, India, and Japan.^1,2 Despite advances in multidisciplinary treatment, the overall prognosis for GBC remains poor. Most patients continue to present at an advanced stage, but in an increasing number of cases, the diagnosis is made after cholecystectomy for suspected benign disease. Current clinical practice and National Comprehensive Cancer Network (NCCN) guidelines recommend additional resection if the primary tumor stage is any greater than T1a. For patients with T1b or higher stage disease, surgical treatment often has evolved to include resection of hepatic segments 4B and 5, porta hepatis lymphadenectomy, and potentially, bile duct (BD) resection performed either primarily or after diagnosis following cholecystectomy.^3,4 Even with surgical clearance, recurrence rates remain high, and some advocate adjuvant chemoradiotherapy.^5,6 In this report, we review outcomes after treatment of GBC and identify prognostic factors that may assist in selecting the extent of the resection and the subpopulation that is likely to benefit from adjuvant therapy.

METHODS

Patients who had carcinoma of the gallbladder diagnosed during the years 2000 through 2010 and who were treated at Oregon Health and Sciences University (OHSU) or Providence Portland Medical Center (PPMC) were retrospectively identified, and their records were reviewed. All the patients had confirmation of GBC by pathology. Those with cholangiocarcinoma or no pathologic diagnosis were excluded. The primary end points included disease-specific survival (DSS) and disease-free survival (DFS). Follow-up included physical examination, history, and a surveillance computed tomographic (CT) scan within 6 months after resection. Additional imaging with a CT or positron emission tomographic (PET) scan was performed at the clinician's discretion. The study was approved by the OHSU and PPMC Institutional Review Boards.

Extent of Resection

The operative surgeon determined the appropriate extent of resection, and no prospective guidelines were used. Any pathology beyond stage T1a was considered to warrant extended resection. Pathologic lymph node (LN) involvement outside the porta hepatis was considered a contraindication for curative resection. Simple cholecystectomy referred to an isolated cholecystectomy via either an open or a laparoscopic approach. Radical cholecystectomy included a segment 4B, 5 hepatectomy. LN dissection (LND) was defined as a yield of 3 or more LNs and generally included removal of the porta hepatis nodes and a variable portion of the nodes in the portacaval space and supraduodenal area (mean nodal yield, 8.42). A major hepatectomy included any hepatectomy larger than a segment 4B, 5 resection. A BD resection included removal of a portion of the extrahepatic biliary tree and enteric reconstruction.

Recurrence

Local recurrence was defined as an enlarging mass on follow-up imaging, with or without biliary obstruction. Distant recurrence was defined as malignant ascites, carcinomatosis, or lymphadenopathy outside the right upper quadrant. A DFS of 0 was applied to patients with gross residual disease or multiple positive margins and disease noted on the first surveillance CT scan. BD involvement was defined as microscopic involvement in a resected BD specimen, gross involvement found at the time of operation, or a positive distal margin of a completely resected cystic duct, typically at the time of reoperation after cholecystectomy. The location and timing of recurrent disease were established from follow-up clinic notes, imaging, or paracentesis for malignant ascites.

Statistical Analysis

Survival analysis was performed with the Kaplan-Meier estimate. Univariate analyses of DFS and DSS were performed by log-rank test. Multivariate analyses were performed via Cox regression. Historically prognostic clinical variables were included in the multivariate analyses. DFS was defined as the interval from resection to identified recurrence, death, or last follow-up. DSS was defined as the time from diagnosis until death or last follow-up. Statistical analysis was performed with SPSS, version 20.

RESULTS

During the 10-year study period, 62 patients with GBC presented to OHSU or PPMC for treatment. Thirty-seven underwent attempted curative resection. Basic demographic and clinical characteristics of the study participants are detailed in Table 1. The median age at diagnosis was 60.5 years; 59% were female. In a high percentage of patients who underwent attempted curative resection (n = 30/37; 81%), GBC had been diagnosed incidentally at a laparoscopic cholecystectomy (LC). In 74% of the patients, at least one LN was examined microscopically.

Table 1.

Demographic and clinical characteristics of patients with resected disease

Parameter	n	%
Age at diagnosis
Median (y)	60.5
Sex
Female	22	59
Male	15	41
Diagnosis
Incidental at cholecystectomy	30	81
Based on symptoms or imaging	7	18
T stage
T1	4	11
T2	13	35
T3	18	48
T4	2	6
Nodal Status
N0	26	70
N1	11	30
Pathologic grade/differentiation
Well, moderate, no comment	27	73
Poor	10	27
Median follow-up (months)	44.2
Recurrence	16	42.8
Local recurrence	12
Distant Recurrence	4
Median DFS (months)	22.6
Median DSS (months)	28.5

Demographic and clinical characteristics of the 25 patients with unresectable disease are reviewed in Table 2. In only 20% of the patients in the unresectable group was GBC diagnosed incidentally. Among those with unresectable disease, 10 had carcinomatosis at exploration, 1 was medically unfit for surgery, 4 had documented LN involvement outside the portal region, and 10 had extensive local disease. Thirteen of 25 patients had jaundice at presentation.

Table 2.

Demographic and clinical characteristics of patients with unresectable disease

Parameter	n	%
Age at diagnosis
Median (y)	65.1
Gender
Female	12	48.1
Male	13	51.9
Diagnosis
Incidental	5	20
Symptomatic	20	80
Jaundice at presentation
Present	10	40
Absent	15	60
Exclusion
Locally advanced	10	40
Carcinomatosis	10	40
Extraportal LN involvement	4	16
Medically unfit	1	4

Table 3 shows the operations performed on the patients with resectable tumors. No uniform or prospective methods were used to determine the operation that any given patient underwent. The volume of liver resection was, at a minimum, undertaken to achieve an R₀ resection if technically feasible. BD resection was pursued if there was any concern or evidence of extension of disease into the main biliary tree. Recurrence was observed in 42.8% (n = 16) of the patients. Twelve of these recurrences were local, and 4 were distant.

Table 3.

Procedures performed

Type	Number
Simple cholecystectomy	5
Simple cholecystectomy, LND	3
Segment 4B,5 hepatectomy, cholecystectomy	5
Segment 4B,5 hepatectomy, cholecystectomy, LND	13
Segment 4B,5 hepatectomy, cholecystectomy, LND, BD resection	8
Major hepatectomy	1
Major hepatectomy, bile duct resection, LND	2

Thirty percent (n = 11/37) of those who underwent resection had pathologic evidence of LN metastasis. Among those with T2 and T3 disease, nodal metastases were present in 36% (n = 5/13) and 33% (n = 6/18), respectively. The presence of LN metastasis was not significantly associated with whether the patient had follow-up chemotherapy or radiation therapy.

Radiation therapy was administered to 16 (43%) of the patients who underwent resection. The patients with positive margins (n = 10; 27%) were significantly more likely to receive radiation therapy than were those with an R₀ resection (likelihood ratio [LR] = 7.7; P = .006). Adjuvant chemotherapy was administered to 14 (38%) of the patients and was not associated with margin or nodal status.

On univariate analysis, BD involvement (n = 9/37) was significantly associated with a lower DFS (P ≤ .001) and DSS (P ≤ .001). T stage was also significantly associated with decreased DFS and DSS (P = .013 and P = .017, respectively). Univariate analysis showed a signification association of margin status with DFS and DSS (P = .017 and P = .019, respectively). Nodal status, grade, chemotherapy, margin status, radiation, and incidental diagnosis were not significantly related to DFS or DSS, nor was BD resection. Multivariate analysis showed a nonsignificant association of LND with DFS and DSS.

Multivariate analysis showed that BD resection (P = .028) and involvement (P = .017) were the only two cofactors significantly associated with DFS (Table 4). In contrast, T stage (P = .026) and BD involvement (P = .049) were significantly related to DSS (Table 5).

Table 4.

Univariate and multivariate analyses: DFS

	Univariate analysis		Multivariate analysis
	HR, 95% CI	P	HR, 95% CI	P
T stage	2.8 (1.4, 5.7)	0.013*	3.4 (0.8, 14.3)	0.09
Nodal status	1.1 (0.4, 3.2)	0.85	0.7 (0.1, 5.3)	0.76
Grade	2.2 (0.8, 6.0)	0.10	1.7 (0.4, 6.2)	0.45
Initial cystic duct involvement	1.2 (0.4, 4.2)	0.79	0.3 (0.1, 2.2)	0.29
Hepatectomy	1.3 (0.4,4.7)	0.63	0.6 (0.1, 4.5)	0.61
LND	0.8 (0.3,2.1)	0.59	0.9 (0.1, 6.1)	0.93
BD resection	0.6 (0.2, 2.2)	0.56	0.1 (0.02, 0.8)	0.028*
Chemotherapy	1.1 (0.4, 3.0)	0.82	0.7 (0.2, 3.0)	0.67
Radiation	1.5 (0.6, 4.0)	0.42	0.9 (0.2, 3.5)	0.85
Positive margin	3.2 (1.2, 8.4)	0.017*	1.4 (0.2, 12.4)	0.77
BD involvement*	6.6 (2.4, 17.9)	<0.001*	15.1 (1.6, 141)	0.017*

DFS in months: no BD involvement: 39.8; BD involvement: 12.7.

^*Statistically significant at P < 0.05.

Table 5.

Univariate and multivariate analyses: DSS

	Univariate analysis		Multivariate analysis
	HR, 95% CI	P	HR, 95% CI	P
T stage	3.5 (1.5, 7.9)	0.017*	7.1 (1.3, 39.5)	0.026*
Nodal status	1.2 (0.4, 3.6)	0.73	0.7 (0.1, 4.7)	0.76
Grade	1.7 (0.6, 4.8)	0.31	1.2 (0.3, 4.4)	0.77
Initial cystic duct involvement	1.1 (0.3, 3.7)	0.94	0.3 (0.1, 1.8)	0.17
Hepatectomy	1.2 (0.3, 4.1)	0.83	0.3 (0.1, 2.3)	0.27
LND	0.8 (0.3, 2.7)	0.79	0.8 (0.1, 6.3)	0.90
BD resection	0.6 (0.2, 2.3)	0.45	0.3 (0.04, 2.1)	0.23
Chemotherapy	0.9 (0.3, 2,7)	0.93	0.2 (0.1, 1.4)	0.11
Radiation	1.7 (0.6, 4.7)	0.32	1.0 (0.2, 4.9)	0.92
Positive margin	3.2 (1.2, 8.8)	0.019*	1.3 (0.2, 10.3)	0.78
BD involvement	5.5 (2.0, 15.6)	<0.001*	8.5 (1.0, 73.6)	0.049*

DSS in months: no BD involvement: 42.0; BD involvement: 26.2.

^*Statistically significant at P < 0.05.

Table 6 shows all the patients with documented BD involvement who underwent resection with curative intent. BD involvement was uniformly fatal, and all patients shown are deceased. Patients 5 and 8 had BD involvement noted on final pathology, whereas intraoperative frozen section analysis showed the BD/distal cystic duct to be free of malignancy. Of the 6 patients with BD involvement in whom GBC was diagnosed incidentally, 3 were documented to have involvement of the distal margin of the initial cystic duct. The sensitivity of pathologic cystic duct assessment following an incidentally diagnosed GBC was 50% (n = 3/6).

Table 6.

Patients with BD involvement

Patient	T	N	Cystic duct	Procedure	Resection	Recurrence	DFS (mo)	DSS (mo)	Adjuvant
1	3	12/15	3	1,2	R1	Distant	24.5	48.3	CRT
2	3	1/1	1	1	R2	Local	0	18.8	CRT
3	4	0/0	3	2,3	R1	Local	0	3.8	XRT
4	3	0/8	1	1,2	R0	Local+distant	17	31.8	None
5	3	1/8	0	1*	R1	Local	3.7	10.4	XRT
6	3	0/8	2	1,2	R0	Local+distant	31.4	51.4	CT
7	3	0/0	3	1	R2	Local	0	5.0	XRT
8	3	0/2	2	1*	R1	Local	0	18.8	CT
9	3	0/2	1	2	R1	Distant	38.1	48.4	CRT

Cystic duct: 0 = not mentioned on pathology; 1 = positive; 2 = identified as negative; 3 = not incidentally diagnosed. Procedure: 1 = 4b,5 segment liver resection; 2 = BD excision; 3 = hepatectomy. Adjuvant: CT = chemotherapy; XRT = radiation therapy; CRT = chemoradiotherapy.

^*Intraoperative frozen sections at the cystic duct/BD were negative. Permanent pathology was positive for invasive cancer.

DISCUSSION

This retrospective analysis of surgically treated GBC from two institutions confirms the aggressive nature of the disease, even in its early stages. The patients with resectable GBC had a notably high rate of incidental diagnosis. Among them, 81% (n = 30/37) had an incidental diagnosis. However, all patients considered, 56% had incidental diagnosis (n = 35/62), similar to the percentage in other publications.^7,8 Incidental diagnosis of GBC is associated with a better outcome, probably because the disease is identified at an earlier stage.⁹ In our population, the trend was toward a better outcome with incidental diagnosis: DFS and DSS had hazard ratios (HRs) of 0.58 (95% confidence interval [CI], 0.19–1.85) and 0.59 (95% CI, 0.16–2.12), respectively. In only 7 of the patients who underwent resection was GBC diagnosed before surgery, and curative resection was attempted in all. It is possible that a larger sample size would have shown an association between incidental diagnosis and DFS or DSS.

Our findings confirm BD involvement as a reliable indicator of poor prognosis (Figure 1). A consistent statistical finding within our population was the significant association of BD involvement with decreased DFS and DSS. This finding persisted across univariate (P < .001 for DFS and DSS) and multivariate (P = .017; P = .049, respectively) analyses. Some institutional and population-based studies have also shown a significant association between BD involvement and outcome,^7,10 although in most, the exact incidence has not been evaluated.^4,11 In addition, in our multivariate analysis, BD resection was significantly associated with DFS (P = .028). In our methods, we defined patients with gross residual disease or multiple positive margins as having a DFS of 0. It follows that after risk adjustment, patients with more extensive resections were more likely to have disease clearance (ie, some duration of DFS). Table 6 details the 9 patients with BD involvement. Four had a DFS of 0, with only 1 of those 4 having undergone BD excision. The association between BD resection and DSS was not significant. Three of the patients with BD involvement received an incidental diagnosis on LC, and pathology reported a cystic duct margin free of disease. In addition, 2 patients had negative cystic duct frozen margins, but final pathology revealed BD involvement.

Figure 1.

Kaplan-Meier DFS stratified by BD involvement. Any involvement: n = 9; no involvement, n = 28.

Accurate assessment of every patient's BD can be challenging, especially after incidental diagnosis and subsequent reoperation. Attention should be paid to the status of the initial cystic duct margin, and caution should be used when interpreting frozen section analysis in a reoperative field. As noted in Table 6, 2 patients in our series had negative intraoperative frozen sections of the cystic duct/BD and ultimately had invasion of the BD. Although BD involvement was uniformly fatal, neither LN status nor adjuvant treatment had prognostic significance in our series.

We found that that of the patients with T2 and T3 tumors, 38% (n = 5/13) and 28% (n = 5/18) had LN metastasis (Table 1), respectively, similar to rates of nodal involvement per stage in other published institutional series of GBC from tertiary referral centers.^4,9,11 Chemotherapy and radiation were administered at similar rates among the patients with N0 and N1 nodal status. Three patients were excluded from surgery because of a positive fine-needle aspiration from a para-aortic LN. Perhaps their exclusion from attempted resection and inclusion in the resectable group of only those patients with nodal disease limited to the periportal region contributed to the lack of effect of LN status on survival. The average number of LNs sampled was 4.3 for N0 patients vs. 6.4 for N1 patients, and the discrepancy may have led to a sampling error, contributing to the lack of effect of nodal status on prognosis. N0 and N1 patients received chemotherapy and radiation therapy at similar rates, and nodal status determined whether patients would benefit from adjuvant therapies. A predictive nomogram based on a retrospective analysis of the SEER-Medicare database predicted that patients with stage T2N0 and higher would benefit from adjuvant chemoradiotherapy (CRT), with patients having the higher stages benefitting the most.⁶ Because of their retrospective and multi-institutional nature, adjuvant treatments were not applied uniformly in our sample. Therefore, it is not possible to judge the efficacy of adjuvant therapy from our data. No randomized, controlled trial has been performed to determine which patients would derive a survival advantage from adjuvant CRT.

Further prospective analysis is necessary for definitive guidance on the use of CRT. An adequately powered, randomized, controlled clinical trial evaluating the benefit of CRT on patients with intermediate-stage (T2N0, T2N1, and T3N0) GBC would be ideal. Currently, the decision to administer postresection adjuvant therapy in patients with GBC is best suited for multidisciplinary evaluation. Factors such as T stage, LN involvement, residual disease, and grade of tumor are all appropriate for consideration. Similar to other institutional reviews of GBC, this study is limited by its retrospective design and relatively small sample size. Statistically robust and generalizable data have been difficult to obtain, in part because of the clinical characteristics of GBC.

However, our results underscore the importance of BD involvement as a significant prognostic marker. It is likely that BD involvement suggests not only local extension of disease, but also a biologically more aggressive neoplasm. Given its prognostic significance, BD involvement may be useful in a prospective study or in a larger series to examine the concordance of cystic duct status after LC, intraoperative frozen section, and final pathology. Regardless of final stage, our data suggest that BD involvement is an appropriate indication for aggressive postsurgical adjuvant therapy.